一种基于稀疏化时序融合的无人物流车障碍物感知方法及系统
By employing a sparse temporal fusion method, combining multi-view images and sparse query sets, and dynamically adjusting computational resources, the real-time and stability issues of obstacle perception in unmanned logistics vehicle scenarios are resolved, achieving efficient obstacle detection and tracking.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HONEYCOMB (WUHAN) MICROSYSTEM TECH CO LTD
- Filing Date
- 2026-04-16
- Publication Date
- 2026-07-17
AI Technical Summary
Existing obstacle perception technologies in unmanned logistics vehicle scenarios involve large computational loads or lack temporal information, making it difficult to operate in real time and stably track dynamic objects. In particular, in low-speed driving and frequent start-stop scenarios, target loss or ID switching is prone to occur.
A sparse temporal fusion method is adopted to obtain BEV feature maps of the current and historical frames through multi-view images. By combining sparse query sets and temporal iterative updates, obstacle position and motion features are optimized, and computing resource allocation is dynamically adjusted to achieve high-precision obstacle perception.
Efficient obstacle detection and tracking were achieved on the edge computing platform, reducing computational load, maintaining stability in occluded scenarios, optimizing system power consumption and latency, and improving perception robustness.
Smart Images
Figure CN122049844B_ABST